Thermal reliability analysis of M3D based on artificial neural network in multi-physics field

MIV(monolithic inter-tier vias) is an important technical means to realize 3D integration because of its ultra-small volume which can provides ultra-high integration density compare to the TSV(through silicon vias). However, in the process of signal transmission, the crosstalk problem in M3D(Monolithic 3D) integrated circuits(ICs) and the heat dissipation and related thermal stress caused by chip density seriously affect the performance of the system. Therefore, this paper considers the MIV structure and uses the finite element simulation method to study its crosstalk coupling effect. Analyze the influence of single factor change of MIV on its crosstalk and temperature stress response under different physical field coupling conditions, then obtain the best influencing factors. Finally, we proposed an M3D array under electro-thermal-mechanical multi-field coupling, analyze the temperature and stress of MIV by COMSOL software. Then we proposed an artificial neural network scheme, the temperature data and stress data of MIV are trained by MATLAB software. The experimental results show that compared with the general BP neural network, the prediction model based on genetic algorithm to optimize the threshold and weight of BP neural network(GABP) has better performance in prediction accuracy and prediction iteration time.